本研究主要目的在建立水庫入流量預測模式,以線性ARMA、SAR;非線性RCA、T AR與非常態模式等時間序列模式,分析曾文水庫入流量紀錄,期能獲得較精確之預測結果,以供相關單住擬定各標的供水量。結果顯示,旬、月入流量之最佳模式為非常態模式。TAR模式旬入流量之預測精度僅低於非常態模式,但TAR模式過於複雜,不滿足模式精簡原則,因此不適用於預測曾文水庫旬入流量。 The major purpose of this research is to establish the forecasting model of inflow for Tsenwen Reservoir by using the linear ARMA, SAR models and nonlinear RCA, TAR, NON-NORMAL models in the field of time series. To expect the result which can be available in making plan for the water quantity of objective supply.The results indicate that the ability of prediction of NON-NORMAL model is the best one for prediction of ten-day and monthly inflow. The accuracy of forecasting for TAR model is only less than NON-NORMAL model, however, TAR model can not be satisfied the parsimony principle. Therefore, TAR model can not be the chosen one.